2007’s in the rearview mirror; 2008’s up ahead
Some of you may be reading this shortly before the end of December, and some after you’ve returned from holiday in January. Either way – please consider this both a review of 2007 and a preview of what I’m predicting for early 2008.
In 2007, member companies achieved some notable milestones including ANSI balloting on VITA 41 (VXS) and VITA 46 (VPX) – destined to be the future of the typical bus-based VME card. Shortly before we went to press, VITA 46.00-2007 (base standard) and VITA 46.1-2007 (signal mapping for VMEbus on VPX) passed ANSI balloting, which clears them as official and formal ANSI specifications.
VPX graduates boot camp
Though VPX products had been introduced to the market this year, this stamp of approval legitimizes the vendor base and guarantees adherence to the standards. In the case of 46.1, it provides a migration pathway between legacy VME and fabric-based VPX products. Recall that 6U VPX deviates from the historical P1/P2/P0 VME DIN connectors and uses six 16-column 7-row MultiGig RT2 connectors and one 8-column 7-row RT2. My supposition is that except for a few military programs earmarking 6U VPX and ushering their support through VSO, the real market activity will be with 3U VPX.
3U is a form-factor size that’s particularly ideal for legacy ATRs and “shoebox” upgrades. 3U VPX provides two 16-column 7-row RT2 connectors plus one 8-column RT2 and 108 pins of multi-gigahertz signal options, compared to CompactPCI’s 75 pins as a system controller. Whereas 3U CompactPCI relies on one connector for the PCI bus, 3U VPX relies on serial fabrics and is sure to be more robust – mechanically and electrically – making it a highly popular choice over 3U CompactPCI, PC/104, or even PICMG’s AdvancedMC-based MicroTCA.
The only Achilles heel for 3U VPX may be price versus the other Small Form Factor (SFF) choices. Surely 3U CompactPCI’s and MicroTCA’s COTS telecom roots will put price pressure on 3U VPX in those applications that are semi-rugged. We’ll have to wait for prices to be announced as products come to market.
In our, there were 40 VPX products listed in late November, and roughly 75 percent of them were 6U. The most visible VPX board vendors are Curtiss-Wright Controls Embedded Computing, Mercury, Extreme Engineering, Micro Memory (now part of VMETRO), TEK Microsystems, and GE Fanuc Intelligent Platforms. Of these, CWCEC seems more focused on 6U for now, but the company recently announced a P.A. Semi-based 3U VPX SBC. At MILCOM last month, GE Fanuc rolled out their Axis family, which included six 3U VPX modules.
2007 was also the year that VITA thumbed its nose atand corporate arm-twisting when it adopted its intellectual property rules. All year the organization garnered headlines as other standards bodies, lawyers, and corporations debated the wisdom of full IP disclosure under the threat of lawsuit. Some VITA members “resigned” as a result of the new policy, stating that there was no way they could possibly disclose all of their IP efforts – either because they simply didn’t know everything going on in their large company, or they just didn’t want to tell anyone.
As the year closes, the ex ante policy seems to be holding up, but there’s no shortage of naysayers anxious to contain this IP wildfire and keep it from spreading to the larger standards organizations. I predict it will prevail; we’ll see how it goes in 2008.
Let’s SWaP for multicore
Asbattle each other over CPU bragging rights – x86 multicore performance will certainly encroach on the PowerPC’s hegemony on the VME platform. I love the PowerPC family, but Freescale’s dual-core 8641 is a literal heater and they’ve got to make a serious attempt to reduce power, or the x86 is going to start driving mission-critical systems and weapons. With virtual machines and a partitioned RTOS/Linux environment, the x86 is looking pretty compelling in traditional mission-critical systems. Come on, Freescale. Show us your roadmap1.
Meanwhile, the P.A. Semi PowerPC has hit a sweet spot in the VME market by allowing designers to choose between performance or power savings with their PWRficient series. Size, Weight, and Power (SWaP) will become a mantra instead of just a differentiator. The days of 40, 60, and 80 W VME boards will become of thing of the past with lower-power multicore CPUs and power-efficient FPGAs that cycle down entire blocks of unused logic. Accordingly, high-octane CPUs like IBM’s nine-core Cell BE will never make the mainstream; instead, they’ll replace racks of SMPs when sufficient shore power is available.
And finally, one more little bombshell hit the shelter as we went to press. This one’ll have to wait until early 2008 to be sorted out. It seems as if VME’s choice of the triple-contact, through-hole pin-and-blade DIN connector was pretty smart. A recent VITA 47 reliability study presented at November’s VSO meeting raised concerns about some surface-mount XMC connectors in really rugged environments. The data hasn’t been made public, but I obtained a copy of a report that showed shearing when these particular surface-mount connectors were run through thermal cycling.
As well, contemporary edge connectors with single contact points are prone to fretting or cyclic fatigue under vibration or thermal stress, and opens may result under a force normal to the contact surface. The RT2 wafer edge connector in VPX did exhibit early signs of fretting and gold wear-through, but that was solved by the wafer shell, thicker plating, and the very rigid keying blocks found on VPX 6U (3) and VPX 3U (2). Certainly in 2008 more cycles will be spent2 studying the mechanical issues revealed as in the recent connector report. VITA has been hinting at a new, even more rugged mezzanine standard proposal for next year – just as XMC cards began hitting the market in 2007. More choice on the way.